Viscosity prediction of refining slag based on machine learning with domain knowledge

Abstract

The viscosity of refining slags plays a critical role in metallurgical processes. However, obtaining accurate viscosity data remains challenging due to the complexities of high-temperature experiments, often relying on empirical models with limited predictive capabilities. This study focuses on the influence of optical basicity on viscosity in CaO–Al₂O₃-based refining slags, leveraging machine learning to address data scarcity and improve prediction accuracy. An automated framework for algorithm integration, parameter tuning, and evaluation ranking framework (Auto-APE) is employed to develop customized data-driven models for various slag systems, including CaO–Al₂O₃–SiO₂, CaO–Al₂O₃–CaF₂, CaO–Al₂O₃–SiO₂–MgO, and CaO–Al₂O₃–SiO₂–MgO–CaF₂. By incorporating optical basicity as a key feature, the models achieve an average validation error of 8.0% to 15.1%, significantly outperforming traditional empirical models. Additionally, symbolic regression is introduced to rapidly construct domain-specific features, such as optical basicity-like descriptors, offering a potential breakthrough in performance prediction for small datasets. This work highlights the critical role of domain-specific knowledge in understanding and predicting viscosity, providing a robust machine learning-based approach for optimizing refining slag properties.

Keywords

refining slag / viscosity prediction / machine learning / domain knowledge

References

[1] Deng ZY, Zhang XM, Hao GY, Wei CX, Zhu MY. Dissolution behavior of Al₂O₃ inclusions into CaO–MgO–SiO₂–Al₂O₃–TiO₂ system ladle slags. Int. J. Miner. Metall. Mater., 2024, 315977

[2] Jönsson PG, Jonsson L, Sichen D. Viscosities of LF slags and their impact on ladle refining. ISIJ Int., 1997, 375484

[3] Chu JH, Bao YP. Mn evaporation and denitrification behaviors of molten Mn steel in the vacuum refining with slag process. Int. J. Miner. Metall. Mater., 2021, 2881288

[4] Zhao LH, Lin L, Wu QF. Experimental study on sulfur removal from ladle furnace refining slag in hot state by blowing air. Int. J. Miner. Metall. Mater., 2016, 23133

[5] Yu ZG, Luo Q, Zhang JY, Chou KC. An insight into the viscosity prediction of ternary alloys with limited solubility. Philos. Mag., 2019, 99192408

[6] Wang C, Chou KC, Yu ZG. Determination of viscosity and surface tension for CaO–SiO₂–CaF₂ slags. J. Solut. Chem., 2020, 496863

[7] Yu ZG, Leng HY, Wang LJ, Chou KC. Computational study on various properties of CaO-Al₂O₃-SiO₂ mold flux. Ceram. Int., 2019, 4567180

[8] X.M. Zha, D.X. Hou, Z.G. Yu, J.Y. Zhang, and K.C. Chou, Comparison of boundary interpolation methods on the geometrical modeling of viscosity for CaO-Al₂O₃-SiO₂ melts, J. Non Cryst. Solids, 562(2021), art. No. 120782.

[9] Yu ZG, Leng HY, Luo Q, Zhang JY, Chou KC. Geometrical modelling of the physicochemical properties of CaO–Al₂O₃–CaF₂ slag at 1873K. Ceram. Int., 2020, 4668075

[10] Yu ZG, Zhang JY, Leng HY, Wu XC, Chou KC. Estimating the density and molar volume of ferrite-based ternary molten slags by geometrical model. Ceram. Int., 2021, 471634

[11] Z.G. Yu, H.Y. Leng, Q. Luo, et al., New insights into ternary geometrical models for material design, Mater. Des., 192(2020), art. No. 108778.

[12] Urbain G. Viscosity estimation of slags. Steel Res., 1987, 583111

[13] Kondratiev A, Jak E. Predicting coal ash slag flow characteristics (viscosity model for the Al₂O₃–CaO–‘FeO’–SiO₂ system). Fuel, 2001, 80141989

[14] Iida T, Sakai H, Kita Y, Murakami K. Equation for estimating viscosities of industrial mold fluxes. High Temp. Mater. Process., 2000, 193–4153

[15] Iida T, Sakai H, Kita Y, Shigeno K. An equation for accurate prediction of the viscosities of blast furnace type slags from chemical composition. ISIJ Int., 2000, 40: S110

[16] Mills KC, Sridhar S. Viscosities of ironmaking and steel-making slags. Ironmaking Steelmaking, 1999, 264262

[17] Ray HS, Pal S. Simple method for theoretical estimation of viscosity of oxide melts using optical basicity. Ironmaking Steelmaking, 2004, 312125

[18] Jiang DW, Zhang JL, Wang ZY, Feng CF, Jiao KX, Xu RZ. A prediction model of blast furnace slag viscosity based on principal component analysis and K-nearest neighbor regression. JOM, 2020, 72113908

[19] Xin ZC, Zhang JS, Lin WHet al.. Sulphide capacity prediction of CaO–SiO₂–MgO–Al₂O₃ slag system by using regularized extreme learning machine. Ironmaking Steelmaking, 2021, 483275

[20] Chen ZW, Wang MH, Meng Z, Wang H, Liu LL, Wang XD. Development of structure-informed artificial neural network for accurately modeling viscosity of multicomponent molten slags. Ceram. Int., 2021, 472130691

[21] Z.Y. Lyu, C. Gu, Z.Y. Lyu, and Y.P. Bao, Composition engineering on the local structure and viscosity of the CaO-SiO₂-Al₂O₃-P₂O₅-FeO slag by machine learning methods, Crystals, 12(2022), No. 10, art. No. 1338.

[22] Liu QS, Wu JJ, Shao YC, Wang H, Zhu X, Liao Q. ANN-based model to predict the viscosity of molten blast furnace slag at high temperatures of > 1600 K. J. Sustain. Metall., 2023, 931020

[23] Y. Liu, J. Wang, B. Xiao, and J.T. Shu, Accelerated development of hard high-entropy alloys with data-driven high-throughput experiments, J. Mater. Inf., 2(2022), art. No. 3.

[24] Y.R. Yuan, Y.D. Sui, P.F. Li, M. Quan, H. Zhou, and A.Y. Jiang, Multi-model integration accelerates Al-Zn-Mg-Cu alloy screening, J. Mater. Inf., 4(2024), art. No. 23.

[25] P.C. Xu, X.B. Ji, M.J. Li, and W.C. Lu, Small data machine learning in materials science, npj Comput. Mater., 9(2023), No. 1, art. No. 42.

[26] Y. Zhang and C. Ling, A strategy to apply machine learning to small datasets in materials science, npj Comput. Mater., 4(2018), No. 1, art. No. 25.

[27] S. Raschka, Model evaluation, model selection, and algorithm selection in machine learning, arXiv preprint arXiv:1811.12808 [2018-11-13]. https://arxiv.org/abs/1811.12808

[28] B. Cao, S. Yang, A.K. Sun, Z.Q. Dong, and T.Y. Zhang, Domain knowledge-guided interpretive machine learning: Formula discovery for the oxidation behavior of ferritic-martensitic steels in supercritical water, J. Mater. Inf., 2(2022), art. No. 4.

[29] J.H. Yuan, Z. Li, Y.J. Yang, A.Y. Yin, W.J. Li, D. Sun, and Q. Wang, Applications of machine learning method in high-performance materials design: A review, J. Mater. Inf., 4(2024), art. No. 14.

[30] Y. Yang, L. Zhao, C.-X. Han, et al., Taking materials dynamics to new extremes using machine learning interatomic potentials, J. Mater. Inf., 1(2021), art. No. 10.

[31] Kim H, Matsuura H, Tsukihashi F, Wang WL, Min DJ, Sohn I. Effect of Al₂O₃ and CaO/SiO₂ on the viscosity of calcium-silicate–based slags containing 10 mass pct MgO. Metall. Mater. Trans. B, 2013, 4415

[32] Cai PC, Luan J, Liu JHet al.. A modified method for calculating the viscosity of multicomponent slags based on Kriging interpolation. Ceram. Int., 2022, 481521803

[33] Pedregosa F, Gaël V, Alexandre Get al.. Scikit - learn: Machine learning in python. J. Mach. Learn. Res., 2011, 12: 2825

[34] Ostroumova L, Gusev G, Vorobev A, Dorogush AV, Gulin A. CatBoost: Unbiased boosting with categorical features. Proc. Adv. Neural Proce. Sys., 2018

[35] Frazier PI. Bayesian optimization. TutOR Oper. Res., 2018, 1: 255

[36] Shahriari B, Swersky K, Wang ZY, Adams RP, de Freitas N. Taking the human out of the loop: A review of Bayesian optimization. Proc. IEEE, 2016, 1041148

[37] Huang GM, Wu LF, Ma Xet al.. Evaluation of CatBoost method for prediction of reference evapotranspiration in humid regions. J. Hydrol., 2019, 574: 1029

[38] Steurer M, Hill RJ, Pfeifer N. Metrics for evaluating the performance of machine learning based automated valuation models. J. Prop. Res., 2021, 38299

[39] Zhang GH, Chou KC. Influence of Al₂O₃/SiO₂ ratio on viscosities of CaO–Al₂O₃–SiO₂ melt. ISIJ Int., 2013, 531177

[40] Machin JS, Yee TB. Viscosity studies of system CaO–MgO–Al₂O₃-SiO₂: II, CaO–Al₂O₃–SiO₂. J. Am. Ceram. Soc., 1948, 317200

[41] Park JH, Min DJ, Song HS. Structural investigation of CaO–Al₂O₃ and CaO–Al₂O₃–CaF₂ slags via Fourier transform infrared spectra. ISIJ Int., 2002, 42138

[42] Li PC, Ning XJ. Effects of MgO/Al₂O₃ ratio and basicity on the viscosities of CaO-MgO-SiO₂-Al₂O₃ slags: Experiments and modeling. Metall. Mater. Trans. B, 2016, 471446

[43] Machin JS, Yee TB, Hanna DL. Viscosity studies of system CaO–MgO–Al₂O₃–SiO₂: III, 35, 45, and 50% SiO₂. J. Am. Ceram. Soc., 1952, 3512322

[44] Forsbacka L, Holappa L, Iida T, Kita Y, Toda Y. Experimental study of viscosities of selected CaO–MgO–Al₂O₃–SiO₂ slags and application of the Iida model. Scand. J. Metall., 2003, 325273

[45] Kim TS, Park JH. Structure-viscosity relationship of low-silica calcium aluminosilicate melts. ISIJ Int., 2014, 5492031

[46] Wu LS, Gran J, Du SC. The effect of calcium fluoride on slag viscosity. Metall. Mater. Trans. B, 2011, 425928

[47] She Y, Ju JT, Jiao ZY, Zhang ZH. Research on the viscosity of CaF₂-SiO₂-Al₂O₃-CaO-MgO slag system by the ther-modynamic software. Adv. Mater. Res, 2013, 634–638: 42

[48] Kozakevitch P. Viscosité et éléments structuraux des aluminosilicates fondus: Laitiers CaO-Al₂O₃-SiO₂ entre 1600 et 2100°C. Metall. Res. Technol., 1960, 572149

[49] Duffy JA. A review of optical basicity and its applications to oxidic systems. Geochim. Cosmochim. Acta, 1993, 57163961

[50] Duffy JA, Ingram MD. An interpretation of glass chemistry in terms of the optical basicity concept. J. Non Cryst. Solids, 1976, 213373

[51] Duffy JA, Ingram MD. Comments on the application of optical basicity to glass. J. Non Cryst. Solids, 1992, 144: 76

[52] M. Cranmer, Interpretable Machine Learning for Science with PySR and Symbolic Regression.jl, arXiv.org [2023-05-02]. https://arxiv.org/abs/2305.01582.

[53] Y.Z. Li, H.Y. Wang, Y. Li, et al., Electron transfer rules of minerals under pressure informed by machine learning, Nat. Commun., 14(2023), No. 1, art. No. 1815.

[54] Wadekar D, Thiele L, Hill JCet al.. The SZ flux-mass (Y–M) relation at low-halo masses: Improvements with symbolic regression and strong constraints on baryonic feedback. Mon. Not. R. Astron. Soc., 2023, 52222628

[55] D. Wadekar, L. Thiele, F.V. Navarro, et al., Augmenting astrophysical scaling relations with machine learning: Application to reducing the Sunyaev-Zeldovich flux-mass scatter, Proc. Natl. Acad. Sci. USA, 120(2023), No. 12, art. No. e2202074120.

[56] P. Lemos, N. Jeffrey, M. Cranmer, S. Ho, and P. Battaglia, Rediscovering orbital mechanics with machine learning, Mach. Learn. Sci. Technol., 4(2023), No. 4, art. No. 045002.

[57] Chartrand P, Pelton AD. The modified quasi-chemical model: Part III. Two sublattices. Metall. Mater. Trans. A, 2001, 3261397

[58] Pelton AD, Chartrand P. The modified quasi-chemical model: Part II. Multicomponent solutions. Metall. Mater. Trans. A, 2001, 3261355

[59] Pelton AD, Chartrand P, Eriksson G. The modified quasi-chemical model: Part IV. Two-sublattice quadruplet approximation. Metall. Mater. Trans. A, 2001, 3261409

[60] Bale CW, Chartrand P, Degterov SAet al.. FactSage thermochemical software and databases. Calphad, 2002, 262189

RIGHTS & PERMISSIONS

University of Science and Technology Beijing